Machine learning optimization for VARTM carbon polymer laminates

Notice

This is an unedited manuscript accepted for publication and provided as an Article in Press for early access at the author’s request. The article will undergo copyediting, typesetting, and galley proof review before final publication. Please be aware that errors may be identified during production that could affect the content. All legal disclaimers of the journal apply.

Year : 2026 | Volume : 14 | 03 | Page :
    By

    M. Rajkumar,

  • Mridula Mavuri,

  • K. Sithananthan,

  • M. Bala Theja,

  • Ankush B. Khansole,

  • Gokul pran. s,

  1. Associate Professor, Department of Information Technology, Sri Krishna College of Engineering and Technology, Coimbatore, Tamil Nadu, India
  2. Instructor, Department of Computer science, Louisiana state university, Shreveport, , Louisiana
  3. Assistant Professor, Department of Mechanical engineering Achariya college of engineering technology, Puducherry, India
  4. Associate Professor, Department of Mechanical Engineering, Santhiram Engineering College (Autonomous), Nandyal, Andhra Pradesh, India
  5. Assistant Professor, Department of Mechanical Engineering, CSMSS Chhatrapati SHAHU college of engineering, Maharashtra, India
  6. Assistant Professor, Department of Computer Science and Engineering, Vel Tech Rangarajan Dr. Sagunthala R&D Institute of Science and Technology, Chennai, Tamil Nadu, India

Abstract

Vacuum-assisted resin transfer moulding (VARTM) is a key low-cost, out-of-autoclave process for manufacturing large-scale carbon-fibre reinforced polymer (CFRP) laminates crucial to aerospace wings, wind-turbine blades, marine hulls, and automotive structures. Unpredictable resin flow often leads to voids, dry spots, and race-tracking defects, resulting in 27.9% scrap rates and lengthy, costly trial-and-error design cycles. Although surrogate models provide rapid impregnation predictions for simple flat-plate geometries, vision-based monitoring is limited to idealized test media without closed-loop integration, and conventional optimization techniques are computationally prohibitive for high-dimensional flow-media networks on realistic aerospace parts, no unified solution yet achieves rapid impregnation prediction, real-time void tracking, and instant optimization. This work proposes a novel hybrid machine-learning framework that integrates a surrogate neural network for thickness-direction impregnation forecasting, a U-Net-based vision system for non-intrusive real-time flow-front and void monitoring, and a proximal policy optimization agent for instant optimal flow-media network generation on new geometries. Trained offline on the VARTM-ML-Opt-2026 dataset from high-fidelity control-volume finite-element simulations and validated on three-dimensionally printed porous media, the closed-loop pipeline achieves 35% fill-time reduction with complete preform saturation and zero trap off, reduces computational cost to 0.74% of full three-dimensional simulations, limits impregnation errors to below 6.5%, and keeps void segmentation errors under 14%. The end-to-end framework offers a scalable, data-driven pathway to defect-free VARTM manufacturing.

Keywords: Machine learning optimization, Surrogate modelling, Real-time void monitoring, Hybrid framework, Real-time void monitoring

How to cite this article:
M. Rajkumar, Mridula Mavuri, K. Sithananthan, M. Bala Theja, Ankush B. Khansole, Gokul pran. s. Machine learning optimization for VARTM carbon polymer laminates. Journal of Polymer & Composites. 2026; 14(03):-.
How to cite this URL:
M. Rajkumar, Mridula Mavuri, K. Sithananthan, M. Bala Theja, Ankush B. Khansole, Gokul pran. s. Machine learning optimization for VARTM carbon polymer laminates. Journal of Polymer & Composites. 2026; 14(03):-. Available from: https://journals.stmjournals.com/jopc/article=2026/view=245101


References

[1] R. Matsuzaki, M. Morikawa, Y. Oikawa, and K. Ushiyama, “Predicting thickness impregnation in a VaRTM resin flow simulation using machine learning,” Composites Part C: Open Access, vol. 5, p. 100158, Jul. 2021, doi: 10.1016/J.JCOMC.2021.100158.
[2] M. Szarski and S. Chauhan, “Instant flow distribution network optimization in liquid composite molding using deep reinforcement learning,” Journal of Intelligent Manufacturing 2022 34:1, vol. 34, no. 1, pp. 197–218, Aug. 2022, doi: 10.1007/S10845-022-01990-5.
[3] D. Wu, R. Larsson, and B. Blinzler, “A preform deformation and resin flow coupled model including the cure kinetics and chemo-rheology for the VARTM process,” International Journal of Material Forming 2020 14:3, vol. 14, no. 3, pp. 421–434, Jul. 2020, doi: 10.1007/S12289-020-01570-Z.
[4] J. M. Jeong et al., “In-situ resin flow monitoring in VaRTM process by using optical frequency domain reflectometry and long-gauge FBG sensors,” Compos. Struct., vol. 282, p. 115034, Feb. 2022, doi: 10.1016/J.COMPSTRUCT.2021.115034.
[5] M. A. Lepore, L. Ferrante, L. Sanguigno, and A. R. Maligno, “A non-crimp fabric mechanical characterization for the production of aerospace components,” Material Design and Processing Communications, vol. 3, no. 5, p. e222, Oct. 2021, doi: 10.1002/MDP2.222;JOURNAL:JOURNAL:25776576;WGROUP:STRING:PUBLICATION.
[6] A. Hindersmann, “Experimental investigation of a method to avoid channel marks during vacuum infusion,” J. Compos. Mater., vol. 54, no. 16, pp. 2147–2158, Jul. 2020, doi: 10.1177/0021998319889120.
[7] S. M. A. Musa, M. H. Dzulkifli, A. I. Azmi, and S. A. Ibrahim, “Embedded and Surface-Mounted Fiber Bragg Grating as a Multiparameter Sensor in Fiber-Reinforced Polymer Composite Materials: A Review,” IEEE Access, vol. 11, pp. 86611–86644, 2023, doi: 10.1109/ACCESS.2023.3304679.
[8] D. Lee, I. Y. Lee, and Y. Bin Park, “Real-time process monitoring and prediction of flow-front in resin transfer molding using electromechanical behavior and generative adversarial network,” Compos. B Eng., vol. 298, p. 112382, Jun. 2025, doi: 10.1016/J.COMPOSITESB.2025.112382.
[9] J. M. Jeong et al., “In-situ resin flow monitoring in VaRTM process by using optical frequency domain reflectometry and long-gauge FBG sensors,” Compos. Struct., vol. 282, p. 115034, Feb. 2022, doi: 10.1016/J.COMPSTRUCT.2021.115034.
[10] D. Wu, R. Larsson, and B. Blinzler, “A preform deformation and resin flow coupled model including the cure kinetics and chemo-rheology for the VARTM process,” International Journal of Material Forming 2020 14:3, vol. 14, no. 3, pp. 421–434, Jul. 2020, doi: 10.1007/S12289-020-01570-Z.
[11] S. Kamath, G. Gandia, N. Adab, E. Mehrdad, D. Qian, and H. Lu, “MACHINE-LEARNING BASED CURING CYCLE OPTIMIZATION IN WIND BLADE MANUFACTURING,” ASME International Mechanical Engineering Congress and Exposition, Proceedings (IMECE), vol. 2, 2024, doi: 10.1115/IMECE2024-146131.
[12] G. Struzziero and A. A. Skordos, “Multi-objective optimization of Resin Infusion,” Advanced Manufacturing: Polymer and Composites Science, vol. 5, no. 1, pp. 17–28, Jan. 2019, doi: 10.1080/20550340.2019.1565648.
[13] J. Mendikute, M. Baskaran, I. Llavori, E. Zugasti, L. Aretxabaleta, and J. Aurrekoetxea, “Predicting the effect of voids generated during RTM on the low-velocity impact behaviour by machine learning-based surrogate models,” Compos. B Eng., vol. 260, p. 110790, Jul. 2023, doi: 10.1016/J.COMPOSITESB.2023.110790.
[14] M. B. Francisco, L. A. de Oliveira, J. L. J. Pereira, A. de Souza, S. S. da Cunha, and G. F. Gomes, “An optimization analysis of a sandwich composite panel with auxetic reentrant core using lichtenberg algorithm based on surrogate modelling,” Mechanics of Advanced Materials and Structures, vol. 31, no. 29, pp. 11953–11967, 2024, doi: 10.1080/15376494.2024.2313165.
[15] Y. Feng, B. Yang, Y. Huang, J. Wang, and P. Causse, “FlowCastNet: A CNN-based surrogate model for the rapid prediction of flow filling patterns in VARTM processes,” Compos. Part A Appl. Sci. Manuf., vol. 204, p. 109656, May 2026, doi: 10.1016/J.COMPOSITESA.2026.109656.
[16] C. Petroll, M. Denk, J. Holtmannspötter, K. Paetzold, and P. Höfer, “Synthetic data generation for deep learning models,” Proceedings of the 32nd Symposium Design for X, DFX 2021, 2021, doi: 10.35199/DFX2021.11.
[17] S. Chen, X. Ouyang, and X. Rao, “Physics-Informed Neural Network (PINNs) for Flow Simulation in Polymer-Assisted Hot Water Flooding,” Processes 2026, Vol. 14, Page 197, vol. 14, no. 2, p. 197, Jan. 2026, doi: 10.3390/PR14020197.
[18] K. D. Humfeld, D. Gu, G. A. Butler, K. Nelson, and N. Zobeiry, “A machine learning framework for real-time inverse modeling and multi-objective process optimization of composites for active manufacturing control,” Compos. B Eng., vol. 223, Oct. 2021, doi: 10.1016/J.COMPOSITESB.2021.109150.
[19] E. Kyriazi et al., “ML-based surrogate cure simulation for predicting process time and temperature overshoot in resin transfer moulding,” Journal of Reinforced Plastics and Composites, 2025, doi: 10.1177/07316844251390926.
[20] M. Shen et al., “PMDI cross-linked rare earth/liquid metal reinforced ANF/MXene membranes for multifunctional electromagnetic interference shielding,” Compos. Part A Appl. Sci. Manuf., vol. 182, p. 108178, Jul. 2024, doi: 10.1016/J.COMPOSITESA.2024.108178.
[21] I. T. Bello et al., “AI-enabled materials discovery for advanced ceramic electrochemical cells,” Energy and AI, vol. 15, Jan. 2024, doi: 10.1016/J.EGYAI.2023.100317.
[22] Y. Guo et al., “Insight into annealing-induced hardening and softening behaviors in a laser powder-bed fusion printed in-situ composite eutectic high-entropy alloy,” Compos. B Eng., vol. 281, p. 111523, Jul. 2024, doi: 10.1016/J.COMPOSITESB.2024.111523.
[23] J. Machado, M. Bodaghi, S. Advani, and N. Correia, “The Development of Data-Driven Algorithms and Models for Monitoring Void Transport in Liquid Composite Molding Using a 3D-Printed Porous Media,” Applied Sciences 2025, Vol. 15, Page 10690, vol. 15, no. 19, p. 10690, Oct. 2025, doi: 10.3390/APP151910690.
[24] J. A. Almazán-Lázaro, E. López-Alba, and F. A. Díaz-Garrido, “Applied computer vision for composite material manufacturing by optimizing the impregnation velocity: An experimental approach,” J. Manuf. Process., vol. 74, pp. 52–62, Feb. 2022, doi: 10.1016/J.JMAPRO.2021.11.063.
[25] J. F. Leon, Y. Li, X. A. Martin, L. Calvet, J. Panadero, and A. A. Juan, “A Hybrid Simulation and Reinforcement Learning Algorithm for Enhancing Efficiency in Warehouse Operations,” Algorithms 2023, Vol. 16, Page 408, vol. 16, no. 9, p. 408, Aug. 2023, doi: 10.3390/A16090408.
[26] F. Grumbach, A. Müller, P. Reusch, and S. Trojahn, “Robustness Prediction in Dynamic Production Processes—A New Surrogate Measure Based on Regression Machine Learning,” Processes 2023, Vol. 11, Page 1267, vol. 11, no. 4, p. 1267, Apr. 2023, doi: 10.3390/PR11041267.
[27] S. H. Kang, H. J. Lee, T. G. Woo, C. S. You, S. H. Lim, and Y. D. Yoon, “Time Delay Implementation in Sensorless Control for Ultra-High-Speed Air Compressor Motor of Fuel-Cell Systems,” 2024 IEEE 10th International Power Electronics and Motion Control Conference, IPEMC 2024 ECCE Asia, pp. 2332–2337, 2024, doi: 10.1109/IPEMC-ECCEASIA60879.2024.10567890.
[28] D. Guinovart, M. S. Chaki, and R. Guinovart-Díaz, “Two-scale asymptotic homogenization analysis of piezoelectric composite materials in generalized curvilinear coordinates,” Compos. B Eng., vol. 284, p. 111677, Sep. 2024, doi: 10.1016/J.COMPOSITESB.2024.111677.
[29] N. Charaf, J. Haase, A. Kulisch, C. Von Elm, and D. Gohringer, “RTASS: a RunTime Adaptable and Scalable System for Network-on-Chip-Based Architectures,” Proceedings – 2023 26th Euromicro Conference on Digital System Design, DSD 2023, pp. 585–592, 2023, doi: 10.1109/DSD60849.2023.00086.
[30] Z. Wang, R. Chu, M. Zhang, X. Wang, and S. Luan, “An Improved Hybrid Highway Traffic Flow Prediction Model Based on Machine Learning,” Sustainability 2020, Vol. 12, Page 8298, vol. 12, no. 20, p. 8298, Oct. 2020, doi: 10.3390/SU12208298.
[31] T. Lavaggi, F. Muhammed, L. Moretti, J. W. Gillespie, and S. G. Advani, “Correlation of the permeability and porosity development of carbon/carbon composites during pyrolysis,” Compos. Part A Appl. Sci. Manuf., vol. 181, p. 108156, Jun. 2024, doi: 10.1016/J.COMPOSITESA.2024.108156.
[32] Y. Zhou et al., “Phosphononitrile based bismaleimide electronic packaging substrate with both fire safety and dielectric properties: Assisting 5G communication,” Compos. B Eng., vol. 280, p. 111489, Jul. 2024, doi: 10.1016/J.COMPOSITESB.2024.111489.
[33] R. Z. Khalid, A. Ullah, A. Khan, A. Khan, and M. H. Inayat, “Comparison of Standalone and Hybrid Machine Learning Models for Prediction of Critical Heat Flux in Vertical Tubes,” Energies 2023, Vol. 16, Page 3182, vol. 16, no. 7, p. 3182, Mar. 2023, doi: 10.3390/EN16073182.
[34] R. Shanthar, R. Prosser, P. Potluri, C. Abeykoon, and S. G. Advani, “Modelling and simulation of the polymer resin flow through fibres during Resin Transfer Moulding: a comprehensive review,” Compos. Part A Appl. Sci. Manuf., vol. 208, p. 109839, Sep. 2026, doi: 10.1016/J.COMPOSITESA.2026.109839.
[35] R. Matsuzaki, M. Morikawa, Y. Oikawa, and K. Ushiyama, “Predicting thickness impregnation in a VaRTM resin flow simulation using machine learning,” Composites Part C: Open Access, vol. 5, p. 100158, Jul. 2021, doi: 10.1016/J.JCOMC.2021.100158.
[36] J. Machado, M. Bodaghi, S. Advani, and N. Correia, “The Development of Data-Driven Algorithms and Models for Monitoring Void Transport in Liquid Composite Molding Using a 3D-Printed Porous Media,” Applied Sciences 2025, Vol. 15, Page 10690, vol. 15, no. 19, p. 10690, Oct. 2025, doi: 10.3390/APP151910690.
[37] M. Szarski and S. Chauhan, “Instant flow distribution network optimization in liquid composite molding using deep reinforcement learning,” Journal of Intelligent Manufacturing 2022 34:1, vol. 34, no. 1, pp. 197–218, Aug. 2022, doi: 10.1007/S10845-022-01990-5.

Ahead of Print Subscription Review Article
Volume 14
03
Received 16/05/2026
Accepted 23/05/2026
Published 26/05/2026
Publication Time 10 Days
1

Login


My IP

PlumX Metrics